In Internet mapping applications, knowing where users are looking at the map can help administrators to make informed decisions about where to update the map or which parts of the map should be optimized for performance. In an effort to understand where people were looking at an online map, Microsoft’s Danyel Fisher created Hotmap, which displays the popularity of map tiles in Microsoft Virtual Earth. You don’t need to look at Hotmap very long to notice which areas are popular. Urban centers, major roads, coastlines, and tourism destinations create a network of the most viewed areas of the map.
(Fisher has published several articles on Hotmap: http://research.microsoft.com/research/pubs/view.aspx?0rc=p&type=Publication&id=1746
http://research.microsoft.com/research/pubs/view.aspx?0rc=p&type=Publication&id=1714
and there is/was an interactive demo available: http://hotmap.msresearch.us/ )
My capstone project idea is to study where people are most likely to look at an online map and then create a predictive model of the “hottest” areas of the map. For a world base map like the one in Virtual Earth, the model might use the above factors as inputs – urban centers/population, roads, coastlines, and points of interest – and would produce a feature class of areas predicted to be the most visited. This feature class could then be used to determine which areas of the map should be updated most frequently, or which areas of the map should be optimized for performance.
There are several ways that the model’s output of “hot areas” could be useful:
· It could act as a guide for an organization to determine which datasets should be most up to date. For example, a GIS Analyst could select images from the organization’s collection that fall within the boundaries of the “hot areas” and mark those for more frequent updates.
· With Internet maps that use cached image tiles for performance, the “hot areas” feature class could act as a guide to which tiles should be pre-created and which should be created on-demand. Creating a tile cache of an entire map, especially at a national scale, can take an extraordinary amount of time and disk space. ArcGIS Server 9.3 will provide the user an opportunity to create just a subset of cache tiles (optionally based on the boundaries of a feature class) and let the server draw the rest on demand as they are visited. The recommended strategy is to pre-create the tiles that will be most heavily-visited, and leave the less-popular tiles to be created on demand. An analyst generating map tiles could configure the caching tools to only create tiles within the “hot areas” feature class. The “hot areas” feature class could also determine which tiles in the map should be updated more frequently.
A capstone project based on this problem could include the following research:
· What techniques have been developed to determine where users look at a map (both printed and online)? Tiled web maps provide enormous potential for studying this, because web server logs can be mined to determine which tiles are the most popular. This is exactly what Microsoft did with Hotmap.
· How do viewing patterns for base maps such as those available in Virtual Earth differ from viewing patterns in more specialized maps that a GIS analyst might create (hydrography maps, cadastral maps, etc.)?
· What geographic trends are visible in Hotmap that could determine the inputs and outputs for a model of “hot areas”? For example, are coastlines generally more popular than surrounding areas?
· How should these geographic trends be represented in the model? For example, should buffered coastlines be included in the model output? By what distance should the coastlines be buffered? Should they only be buffered on one side or do people also look at the ocean next to the coast? etc.
· Would the model output need to be generalized for specific purposes? For example, the ArcGIS Server tile caching tools perform better with a few large input features instead of numerous small features. I don’t even know if they support polygons with holes. Etc.
· Is this model applicable in other geographic entities besides the US? Do any parts of the model need to be adjusted when applying it to cities, counties, other countries, etc. ? Can the model be abstracted in order to be useful to everybody?
· In what ways (besides the two mentioned previously) would this model output be useful?
These are questions that might be helpful to discuss with a faculty advisor to help guide my research. I have technical background in ArcGIS Server map caching- I have presented on this subject at several ESRI conferences. I probably have enough experience with ESRI geoprocessing tools to create the model. I think I just need an advisor with an interest in this area who can help me focus my research to the most important of the above points.
I have visited with Clint Brown, director of the Software Products department at ESRI, and received his support for this project idea. If possible I would also like to visit with Danyel Fisher of Microsoft, who created Hotmap, during some point of my research.
My biggest concern about this project idea is if it will fulfill the required scope of an MGIS capstone project. I would like to receive more information from you about the required deliverables for the project and the research process.